摘要
从改进传统的模式识别判别函数法的目的出发,提出一个基于统计学和凸二次规划的模式识别方法,简称LCL方法.文中对众所周知的关于植物分类的IRIS数据进行计算比较,若IRIS中的150个例子全部参加统计、学习,则正态分布Bayes判别函数法的正确分类率为80%,而采用本文LCL方法,取每类前30个例子作为典型正例,后20个例子作为一般正例,则所确定的判别函数的正确分类率达99.3%,表明了该方法的有效性和优越性.
A pattern recognition method based on statistics and convex quadratic programming, which is called LC'L method for short, is presented to improve the tranditional decision functions of pattern recognition. The LCL method is compared with Bayes method in normal distribution on IRIS data in the plant classification: the correct classification rate is 80% for Bayes method while 99.3% for the LCL method. The results show that the LCL method is efficient, and advantageous.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
1996年第4期311-316,共6页
Pattern Recognition and Artificial Intelligence
基金
福建省自然科学基金
关键词
模式识别
凸二次规划
统计学
判别函数法
Pattern Recognition, Convex Quadratic Programming, Statistics, Decision Function Method, Machine Learning.
